

Before you begin the troubleshooting process, ensure the suggested infrastructure specifications located in the System Requirements guide are met.
Consider the following components that impact performance in the new HTML Viewer.
Before you begin the troubleshooting process, familiarize yourself with the following terminology.
The asynchronous nature of document conversion and viewing means there are several components to check in order to identify performance issues. As a rule, determine a method to test suspected performance problems, eliminating as much uncertainty as possible, before you make any modifications.
The following section assists you in investigating large parts of the infrastructure. You don’t need to follow this section linearly.
First, determine whether the conversion process itself is causing problems or whether the problem lies in the client, document viewer services, agents, web, or SQL Servers. By identifying the cause of the problem, you can narrow down the performance issue to the fewest components.
If documents take many seconds to convert and download, use the following steps to pinpoint the source of the issue:
If the page refresh is faster than the original page load, there is a problem with the conversion process. See Conversion is a problem.
If the page refresh took the same amount of time as the original page load, conversion is not the source of performance issues. See Conversion is not the problem.
If you determined conversion is a problem, ensure you narrow down the conversion problem to one of the following performance components:
If you determined the conversion agent process takes a long time, you can run the Conversion time metrics gathering.sql script available in the
The second set of output separates conversion times by percentiles. In the following example 95% of all image conversions took 4.473 seconds or less to complete; 90% of all native conversions took 0.706 seconds or less to complete:
You can determine whether the conversion agent is the source of performance issues in your environment by reviewing the two outputs from running the “Conversion time metrics gathering.sql” script. While conversion times are dependent on your hardware and document set, the following are general guidelines for the average conversion process:
The conversion agent is not the source of the performance issues if conversion times are significantly lower than the time required to initially load a document. If conversion times are comparable to the time required to initially load a document, the conversion agent is the primary source of performance issues.
If you determine the conversion agent is the source of your performance issues, consider the following performance solutions:
The last two options assist with perceived performance rather than improving actual performance. For either of the last two options, test to ensure the environment can handle the load and that more conversion agents aren't needed.
If the web server is the source of performance issues, Relativity runs slowly and the web server likely sits at 100% CPU and/or RAM utilization. This means the web server doesn’t have adequate hardware to support the environment.
If the document viewer service is the source of performance issues, Relativity API runs slowly. This means the web server doesn’t have adequate hardware to support the environment.
Possible solutions:
If the conversion complete agent is the source of performance issues:
If you determine the web server is the source of your performance issues, consider the following performance solution. Use 4 CPU cores and 4GB of RAM for every 25 active users on a web server in Relativity 9.0 and above. This is significantly higher than what was required in earlier versions. It’s possible if your environment had no previous performance concerns, after an upgrade to Relativity 9.0 or higher, your environment exhibits performance issues.
If conversion is a problem, but the document viewer service, conversion agents, and conversion complete agents and web servers all perform as expected, then consider environmental scaling as the source of performance issues. To confirm that environmental scaling is the root cause, examine the "conversions_rp<Resource Pool ID>" topic's subscriptions. If large numbers of messages are piling up on these subscriptions, then add more conversion agents to the resource pool. Examine "conversions_responses" if any of those subscriptions have large number of messages piling up on these, then add more conversion complete agents.
As a base recommendation dedicate one conversion agent solely to conversion for every 100 concurrent reviewers and 1 conversion complete agent per web server on the core agents server. However, this ratio is only a suggestion, as different workflows may require a different number of conversion agents. For example, if your workflow has many documents taking minutes to convert, additional conversion agents may be recommended.
If conversion is not the source of the performance issues, the environment is the likely source.
Before sending a document to the client machine, the web server needs to know where the document resides on the disk.
If the entire environment is experiencing poor performance, Relativity SQL is not the source. If Relativity SQL is the source, performance issues occur only in a single workspace or workspaces using the same template.
This script is useful for identifying SQL problems. There is a low chance that there is locking on the ConvertedCacheFile table, which suspends queries accessing the cache location. If you see many suspended queries accessing the ConvertedCacheFile table with long lock waits when you check the output of this script, then the table has a locking problem.
If you determine Relativity SQL is the source of your performance issues, consider the following performance solutions:
If Relativity SQL isn’t the source of your performance issues, it is likely the web and cache servers are the source.
To check your web and cache servers’ performance, use the following procedure:
Zoom in on the segment containing /Relativity/CustomPages/C9E4322E… and /Relativity/CustomPages/5725cab5… (Columns are rearranged in the screen shot.) These calls receive the CacheID for the document (C9E) and download the converted file (5725).
The web and cache servers respond quickly if “taken” times are small. However, if “taken” times are in seconds or hundreds of milliseconds, sending data from the servers to the client is the source of your performance problem.
If you determine the web or cache server is the source of your performance issues, consider the following performance solutions:
If you determine the Client machine is the source of your performance issues check to see if the performance problem is present on a different machine running the same browser.
If you determine the web browser is the source of your performance issues, check to see if the performance problem is present in Chrome, Firefox, and Internet Explorer 10/11.
Testing shows some performance problems exist only in Internet Explorer 10. If Internet Explorer is preferred, Internet Explorer 11 could provide better performance.
If you determine metadata is the source of your performance issues, check to see if the performance problems are present when there are no persistent highlight sets or markup sets. This is identical to the troubleshooting done for the ActiveX viewer.
If you determine data is the source of your performance issues, check to see if the performance problem is present on different documents and document types. Large (50+ pages) image files/productions, spreadsheets, and some other file types exhibit some performance issues in the HTML viewer.
You can also use knowledgebase articles to diagnose and solve viewer performance issues not explained in this guide. The knowledgebase articles are provided by Customer Support and are available via the Relativity Community.
Why was this not helpful?
Check one that applies.
Thank you for your feedback.
Want to tell us more?
Great!